COVI-19 Imaging for AI
Community AI for digital triage developed by patients, clinicians, and data scientists to fight COVID-19!
These two chest x-ray (CXR) images are from a 72-year-old woman who has a cough and respiratory distress from last year (left) and now (source). The yellow circle and ovoid indicate the typical subpleural peripheral opacities.
Digital Triage
Chest x-ray is a cheap scaleable imaging method to diagnose COVID-19 infections. The decisions made off of these digital radiology screens may influence patient placement, quarantine, ICU admission, and ventilator support. AI trained to improve and automate COVID radiology interpretation is digital triage to fight COVID-19 spread on the front lines.
Clinical AI
China deployed clinical-grade AI trained on massive state-run datastores of radiographs as part of their national strategy to halt the COVID-19 epidemic. While the US cannot duplicate such centralized efforts because of its largely privatized health-care system,. patients have the right to securely share their private clinical data with researchers to develop clinical-grade AI.
Data Philanthropy
Without a nationalized health infrastructure, data philanthropy in the US will be critical to the realization of clinical AI. Since the 1980s, HIPAA has mandated that patients can share their clinical data for research. We are building an AI Community where patients donate their imaging to our COIVID Imaging Repository for clinicians and data scientists to build clinical-grade AI.
Digital Triage
CXR is the first imaging method to diagnose COVID-19 coronavirus infection in Spain. Patients with very high suspicion and confirmed infection are examined with portable radiology equipment. The decisions made off of these radiology screens may influence patient placement, quarantine, ICU admission, and ventilator support. Therefore, improving the precision of radiology interpretation with AI may help fight COVID-19 spread on the front lines.
Clinical AI Community
If it takes a village to raise a child, it takes a whole community to train clinical-grade AI. Accurate AI development is reliant on well-annotated training datasets that clinicians should ideally curate. We have developed a HIPAA-compliant web application to allow patients to donate their imaging, clincians to curate the imaging for COVID-19 features, and data scientists can leverage this curated data to train clinical AI. We are submitting our app for IRB approval and once we get it this is how it should work.
Patients
Patients sign-in to our app to answer secure surveys about their COVID-19 experience including where and when they had their imaging performed. Our partners securely transfer de-identified electronic health records to our CovidImaging Repository. Patients can access their imaging through the app and learn about our AI research and insights.
Clinicians
Clinicians sign-up to our app to curate the data that patients donate. Such curation may include expert segmentation of COVID features on X-ray or CT scan. We also align clinical features from orthologouts data sources such as SickWeather that can label individual patients for machine learning.
Data scientists
Data scientists will access deidentified patient imaging with high quality clinical curation. Data scientists can use these crowd-sourced labeled imagesets to train, test and validate clinical AI models. AI development will be iterative and closely guided by clinical curation.
A neural network can help spot Covid-19 in chest x-rays
March 25, 2020
First real open source CNN and dataset to detect COVID-19 on chest x ray. This may be useful for digital triage to better manage the pandemic on the front lines. Manuscript includes a GitHub repo with over 5K CXRs.
Behold.AI among the first to IDs abnormal chest X-rays from COVID-19 patients
March 24, 2020
AI-powered cognitive computing platform vendor behold.ai today announced that its artificial intelligence-based red dot algorithm quickly identifies chest X-rays from COVID-19 patients as abnormal.
"Out of 28 X-rays reviewed from patients with COVID-19, the vendor reported it correctly identified 85% of them as abnormal using red dot."
February 26, 2020
Software that reads CT lung scans had been used primarily to detect cancer. Now it's retooled to look for signs of pneumonia caused by coronavirus.
February 19, 2020
Alibaba claims its new system can detect coronavirus in CT scans of patients’ chests with 96% accuracy against viral pneumonia cases. And it only takes 20 seconds for the AI to make a determination – according to the report, humans generally take about 15 minutes to diagnose the illness as there can be upwards of 300 images to evaluate.
COVID Imaging Repository
Countries like China have trained AI on state-curated centralized data-stores of radiographic imaging as part of a strategy to effectively halt their COVID-19 epidemics. In the US, however, privacy regulations, lack of interoperability and active health information blocking has curtailed the availability of data to train clinical grade AI. To overcome these obstacles, we are crowdsourcing an open COVID-19 Imaging Repository for AI. Using a HIPAA-compliant web application, patients will be able to access their clinical imaging and may share their deidentified data with any IRB-approved study including our own university research. Data scientists will use patients’ anonymized data to develop and deploy AI for digital triage of COVID-19. Please sign up for future updates including how to contribute to the COVID Imaging Repository research study.
Executive profiles
We bring together the best team from academia and industry across artificial intelligence, epidemiology. technology, and disaster mitigation to leading fight COVID-19!
Dexter Hadley, MD/PhD
Chief of Artificial Intelligence & Assistant Professor
UCF College of Medicine, Department of Clinical Sciences
Dr. Hadley serves as an Assistant Professor of Pathology at UCF College of Medicine in the Department of Clinical Sciences. As a new faculty member, he is also the founding Chief of a new Division of Artificial Intelligence to leverage the transformative power of AI from medical school through clinical practice. He is also the PI of The Hadley Laboratory and has repeatedly been funded by the NIH to crowd-source large-scale data and curation for developing clinical artificial intelligence. In general, the end point of his work is rapid proofs of concept clinical trials in humans that translate into better patient outcomes and reduced morbidity and mortality across the spectrum of disease.
Elena Cyrus, MPH, PhD
Assistant Professor,
UCF College of Medicine, Department of Population Health Sciences
Dr. Cyrus will be joining the faculty of UCF in June as an expert global health epidemiologist with over fifteen years of cross-cultural clinical trial management experience in Sub-Saharan Africa, India, Latin America and the Caribbean (LAC), and, the U.S. She trained at Yale School of Public Health as an HIV epidemiologist where she received the prestigious Fogarty Global Health Fellow for her work. Dr. Cyrus has substantial experience assessing the public health needs of of various minority and at-risk populations, and she has repeatedly been funded by the NIH to devise and implement interventions to address the public health needs of at-risk populations.
David Metcalf, PhD
Director of Mixed Emerging Technology Integration Lab (METIL) & Professor
UCF Institute for Simulation and Training.
Dr. Metcalf has more than 20 years' experience in the design and research of web-based and mobile technologies converging to enable learning and healthcare. He serves as the Director of the Mixed Emerging Technology Integration Lab (METIL) at UCF's Institute for Simulation and Training. He has written the book on Blockchain in Healthcare, and his team has built mHealth solutions, simulations, games, eLearning, mobile and enterprise IT systems for Google, J&J, VA, U.S. military and UCF's College of Medicine among others. Dr. Metcalf frequently presents at industry and research events shaping business strategy and use of technology to improve learning, health and human performance.
Eric Kant
Disaster Management Consultant
Kant Consulting Group, LLC
Mr. Kant is a former firefighter, paramedic and emergency manager with a proven history of saving lives with innovation, applying operational expertise, and offering hands-on guidance at significant events and disasters worldwide. He serves as the disaster portal coordinator for the International Association of Emergency Mangers and works as an innovation entrepreneur. He has been recognized by the Department of Defense, National Geospatial Intelligence Agency, NATO, DHS S&T and others for innovative thinking and applying expert analysis to complex cascading operational interdependencies. During his career, Mr. Kant has supported real-time command/control operations for multiple agencies around the world, including the Florida Night of Tornadoes, the World Trade Center Disaster, hurricanes Katrina and Sandy, and many other significant events and disasters over the past two decades.
Trainee and Staff Profiles
We've got the best students, postdocs and staff behind the scenes working to make an impact.
Olga Petan
Chief AI Engineer
Ms. Petan builds end-to-end machine learning-based medical diagnosis systems with open source libraries compliant with HIPAA regulations. Her goal is to help alleviate some of healthcare's biggest bottlenecks and obstacles - by building infrastructure and tailored solutions that will help doctors and medical professionals reach more accurate diagnosis and treat more patients. She is aMicrosoft certified in Azure and Big Data analytics.
Dmytro Lituiev, PhD
AI Specialist
Bakar Computational Health Sciences Institute, UCSF
Dr. Lituiev is an expert in deep learning. He is the only certified NVIDIA Deep Learning Ambassador at UCSF, and he also teaches advanced data science courses at UC Berkeley. He hosts the weekly SF Deep Learning Study Group at UCSF where a community of deep learning practitioners across SF Bay Area, from both academia and industry, meet to review state-of-the-art DL approaches and tackle novel projects. He has earned a PhD in computational biology, and his passion is to develop machine learning and visualization tools, for Life Sciences, and other applications. Dr, Lituiev strives to turn demands of experimentalists into automatized solutions by applying statistical, image and signal processing and statistical tools.
© 2020